On Learning: by Oliver G. Selfridge, Chief Scientist of the Fundamental Research Laboratory, General Telephone & Electronics, Waltham, MA
(see also Wikipedia at his name.)

Source: (Lawler) Cognitive Science and Education: the future within our reach:
a series of Invited Talks at Purdue University, 1988-1989
This panel is linked directly from “Mentors” in the right sidebar and from the Introductions page, LC0aV, which serves as a directory to all these talks.

Selfridge Abstract: On Learning: the challenge of AI
Pointing to the ancient roots of AI, Selfridge highlights the links between Cybernetcs and modern AI, where his studies with Wiener and long friendship
with McCulloch are a bridge, as is his articulate formulation of behavior in terms of control theory. His examples of increasing complexity lead to the
challenging formulation of the central issue for making sense of learning: “how can we represent the world so that learning is easy?”
Oliver gently corrects Lawler’s faux pas by emphasizing the deep reason Minsky valued the McCulloch-Pitts work: they established that
neural nets could compute anything a Turing machine could compute and thus, that the brain is some kind of computer.

Introduction by Bob Lawler, 1/12/1989, 27mb

What I think about Learning, 32mb

The early days of Artificial Intelligence (AI),34mb

Learning: always a central theme in AI, 35mb

Learning seen as adaptation with control loops, 43mb

Adaptation & Learning in living systems: eg.e-coli, 30mb

Control in more complex life forms: mating of moths, 34mb

How do we represent the world so learning is easy? 30mb

We learn because we have a rich multiplicity of purposes,37

The complexity of purposes: the Dandelion Caper, 21mb

Print Friendly, PDF & Email